- Messages
- 3,992
- Likes
- 1,873
- Points
- 2,130
A single photo is enough to unearth a person's entire digital life.
A photo of a face can be enough to trigger an automated search across almost a person's entire open digital life. Clearview AI has created and tested a prototype of InquiryIQ, which can autonomously crawl websites, examine photos, and compile a dossier containing social media accounts, addresses, phone numbers, places of work, aliases, and connections. The development has not yet been handed over to the police and is not yet ready for release in its current form, but InquiryIQ demonstrates how quickly AI can transform a standard facial search into an automated digital investigation.
Traces of InquiryIQ were discovered in files that the Clearview AI login page sent to visitors' browsers before authorization. Thousands of interface lines, instructions, warnings, and function descriptions remain in the open source code. The files do not reveal the server side and do not allow for testing the system's actual effectiveness, but they do provide a detailed description of the intended mechanics.
The starting point is a standard search through Clearview AI's facial recognition system. The investigator selects the information found, after which InquiryIQ expands the search automatically. Age, gender, race, hair color, and eye color can be added to the profile. The interface clearly states that these characteristics help the AI make more accurate decisions for subsequent queries.
The program then launches a search of websites and images, opens pages, and applies facial recognition to the found photos. The results gradually form a Candidate Graph, a graph of possible identities and connections. A person's profile can include probable addresses, phone numbers, employers, social media accounts, arrest records, aliases, and suspected acquaintances. In essence, InquiryIQ automates a significant portion of OSINT that investigators previously had to conduct manually across numerous individual sources.
The uniqueness of this approach lies not so much in access to new information as in the speed with which it connects previously published fragments. An old photo on an event website, a forgotten social media profile, a mention of an employer, and a photo with another person can lie scattered across the internet for years. InquiryIQ attempts to automatically connect such traces to a single individual. Work that would take analysts days or weeks is potentially reduced to minutes.
Clearview AI incorporated various language models for its experiments. The prototype interface included models from xAI and Amazon Bedrock. Clearview AI CEO Amos Kyler explained that the choice of models was intended for internal comparison by developers, not for police officers. According to Kyler, no law enforcement users have used InquiryIQ. The company also claims that it has not yet determined the suitability of any specific language model for such a tool.
Clearview AI already has the foundation for its new search engine. The company claims to have a database of over 70 billion facial images collected from publicly available internet sources, including news sites, public social media pages, and websites with mugshots of detainees. The platform is designed for law enforcement, government, intelligence, and other governmental agencies. Clearview AI specifically states that it does not obtain data from closed or protected accounts.
The transition from a lookalike search to the automatic construction of a detailed profile significantly increases the risk of errors. The InquiryIQ interface itself warns that automatically found demographic information, accounts, and arrest records may be inaccurate. Before adding a result to a profile, an employee must confirm independent verification of the information. Clearview AI insists that the system should generate leads for further work, while the final decision rests with humans.
The problem is that generative models do not always reproduce the same chain of reasoning and can incorrectly associate information about different people. Automation also makes such checks much cheaper: investigators can examine far more people without lengthy manual work. Privacy experts fear that the elimination of time and labor constraints will significantly expand the practical capabilities of mass searches of open data.
Clearview AI has already faced significant controversy over its handling of biometric data. In 2024, the Dutch regulator fined the company €30.5 million for GDPR violations, including processing biometric data without a legal basis. In the US, a 2022 court settlement limited the provision of the database to private organizations, while preserving the ability of government agencies to use the service, subject to existing restrictions.
Automated investigations also have a potential advantage. If the system stores queries, models used, pages found, and navigation sequences, the investigator's actions can be reconstructed much more accurately than a typical manual search. However, such transparency will only come with full logging of AI work. For now, InquiryIQ remains an experimental prototype, and Clearview AI has not announced a launch date.
A photo of a face can be enough to trigger an automated search across almost a person's entire open digital life. Clearview AI has created and tested a prototype of InquiryIQ, which can autonomously crawl websites, examine photos, and compile a dossier containing social media accounts, addresses, phone numbers, places of work, aliases, and connections. The development has not yet been handed over to the police and is not yet ready for release in its current form, but InquiryIQ demonstrates how quickly AI can transform a standard facial search into an automated digital investigation.
Traces of InquiryIQ were discovered in files that the Clearview AI login page sent to visitors' browsers before authorization. Thousands of interface lines, instructions, warnings, and function descriptions remain in the open source code. The files do not reveal the server side and do not allow for testing the system's actual effectiveness, but they do provide a detailed description of the intended mechanics.
The starting point is a standard search through Clearview AI's facial recognition system. The investigator selects the information found, after which InquiryIQ expands the search automatically. Age, gender, race, hair color, and eye color can be added to the profile. The interface clearly states that these characteristics help the AI make more accurate decisions for subsequent queries.
The program then launches a search of websites and images, opens pages, and applies facial recognition to the found photos. The results gradually form a Candidate Graph, a graph of possible identities and connections. A person's profile can include probable addresses, phone numbers, employers, social media accounts, arrest records, aliases, and suspected acquaintances. In essence, InquiryIQ automates a significant portion of OSINT that investigators previously had to conduct manually across numerous individual sources.
The uniqueness of this approach lies not so much in access to new information as in the speed with which it connects previously published fragments. An old photo on an event website, a forgotten social media profile, a mention of an employer, and a photo with another person can lie scattered across the internet for years. InquiryIQ attempts to automatically connect such traces to a single individual. Work that would take analysts days or weeks is potentially reduced to minutes.
Clearview AI incorporated various language models for its experiments. The prototype interface included models from xAI and Amazon Bedrock. Clearview AI CEO Amos Kyler explained that the choice of models was intended for internal comparison by developers, not for police officers. According to Kyler, no law enforcement users have used InquiryIQ. The company also claims that it has not yet determined the suitability of any specific language model for such a tool.
Clearview AI already has the foundation for its new search engine. The company claims to have a database of over 70 billion facial images collected from publicly available internet sources, including news sites, public social media pages, and websites with mugshots of detainees. The platform is designed for law enforcement, government, intelligence, and other governmental agencies. Clearview AI specifically states that it does not obtain data from closed or protected accounts.
The transition from a lookalike search to the automatic construction of a detailed profile significantly increases the risk of errors. The InquiryIQ interface itself warns that automatically found demographic information, accounts, and arrest records may be inaccurate. Before adding a result to a profile, an employee must confirm independent verification of the information. Clearview AI insists that the system should generate leads for further work, while the final decision rests with humans.
The problem is that generative models do not always reproduce the same chain of reasoning and can incorrectly associate information about different people. Automation also makes such checks much cheaper: investigators can examine far more people without lengthy manual work. Privacy experts fear that the elimination of time and labor constraints will significantly expand the practical capabilities of mass searches of open data.
Clearview AI has already faced significant controversy over its handling of biometric data. In 2024, the Dutch regulator fined the company €30.5 million for GDPR violations, including processing biometric data without a legal basis. In the US, a 2022 court settlement limited the provision of the database to private organizations, while preserving the ability of government agencies to use the service, subject to existing restrictions.
Automated investigations also have a potential advantage. If the system stores queries, models used, pages found, and navigation sequences, the investigator's actions can be reconstructed much more accurately than a typical manual search. However, such transparency will only come with full logging of AI work. For now, InquiryIQ remains an experimental prototype, and Clearview AI has not announced a launch date.